mirror of
https://git.freebsd.org/ports.git
synced 2025-04-29 01:56:37 -04:00
bcd2014823
Remove the graphics/libpng dependency and configure option. PNG support is now entirely built-in. graphics/libGLU is only needed to build the glviewer test program. Its presence on the host system doesn't affect the installation. Remove reference to removed port x11-toolkits/fox14. Pet portclippy(1) and portlint (1). Update patches due to changes in base and to make them more upstream- friendly. Changes: - New PNG image support. First, its faster! Also, it no longer requires libpng; the PNG format support is now built-in, although the libz compression library is still required [for now...]. - fxsavePNG() supports some fine-control via new save-flags. You can analyze the image, and shrink output by taking advantage of image features; for example, saving opaque image means alpha-channel may be dropped. If the image is gray-levels only, the RGB may be dropped to plain gray only. Also, shrinking further by emitting colormapped [indexed-color] image is possible if only a few colors are used. Finally, pre-compression filtering, as well as desired compression level can be selected. - PNG I/O is optimized with when target ISA x86-64-v2 or higher are selected. - FXStat::isSame() checks if two files are the same (same inode). - FOX Desktop Calculator augmented with Unicode button labels (please select a font that has the math symbols!), also now supports additional functions, common physics constants, and other features. - The fxCPUFeatures now can detect AVX512 presence. - New FXPerformanceCounter and associated macros may be used to count clockcyles of critical code segments. - FXAtomic.h APIs now mostly inlined for lower overhead. - Read processor ticks on AARCH64. - Adie text editor undo buffer size and undo buffer items can now be configured. - New QOIF (Quite OK Image Format) now supported for either images (FXQOIFImage) or icons (FXQOIFIcon). - FXColor to/from FXVec3d, FXVec3f, FXVec4d, FXVec4f now using SSE if compiled for x86-64-v2 or higher. - Updated byte swap APIs in fxendian.h. - New APIs in FXMat3f, FXMat3d, etc. classes to set up mirror-matrix. - Moved new hash32() etc. functions into fxendian.h. - Bug fix in FXIODevice reading > 1GB files in one readBlock(), writeBlock() call. - Markdown syntax coloring in Adie. - Per-syntax mode setting for removing trailing spaces in Adie saving a text file. - Subtle change in operation of FXPath::relative(). - Support for CRC32 calculations added. - Additional conversion options in FXString::fromFloat() and FXString::fromDouble(): thousands groupings, force decimal point, and hexadecimal float output. - FXJSON JSON loader now may report duplicate key warning reading json file. - FXMappedFile improvements. - FXString unicode escapes bug fix.
17 lines
747 B
C++
17 lines
747 B
C++
pthread_rwlockattr_setkind_np() is not implemented on FreeBSD. This function
|
|
was contained in an ifdef in version 1.7.75. The release notes for 1.7.76 only
|
|
state that not all systems have this function, but the supposed fix is a
|
|
regression on FreeBSD.
|
|
|
|
--- lib/FXReadWriteLock.cpp.orig 2021-04-02 00:40:37 UTC
|
|
+++ lib/FXReadWriteLock.cpp
|
|
@@ -70,7 +70,9 @@ FXReadWriteLock::FXReadWriteLock(){
|
|
FXASSERT_STATIC(sizeof(data)>=sizeof(pthread_rwlock_t));
|
|
pthread_rwlockattr_t rwlockatt;
|
|
pthread_rwlockattr_init(&rwlockatt);
|
|
+#if defined(__GLIBC__)
|
|
pthread_rwlockattr_setkind_np(&rwlockatt,PTHREAD_RWLOCK_PREFER_WRITER_NP);
|
|
+#endif
|
|
pthread_rwlock_init((pthread_rwlock_t*)data,&rwlockatt);
|
|
pthread_rwlockattr_destroy(&rwlockatt);
|
|
#else
|